Optimal inspection policies for standby systems

Abstract
An auxiliary unit is normally idle or in cold standby. If a demand for the unit's service occurs, the unit must be available to satisfy it, or else “catastrophe” occurs. Discrete time Markov decision process models are formulated and policies for periodic inspection and maintenance of such a unit are derived that maximize the expected time until a catastrophe occurs. The models recognize that inspection, maintenance, and repair periods are of non-zero duration during which the unit is vulnerable. They also account for the possibility of hazardous inspection that may damage the unit and various forms of imperfect repair